[docs] fix: correct Qwen-Image 2.1 model size

.agents/knowledge/topics/fix_patterns.md

@@ -797,6 +797,18 @@ Based on the fix type, write the fix entry to the appropriate document:
   dropping a supported Python version.
 - **Related Constraint**: N/A
 
+### Versioned model rows require independent size verification
+- **Date**: 2026-09-21
+- **Symptom**: The supported-model table listed Qwen-Image 2.1 as 20B even though its official
+  model card identifies the visual generation component as 7B.
+- **Root Cause**: The new version inherited the parameter count of earlier Qwen-Image entries
+  without independently verifying the changed architecture.
+- **Fix**: Correct the README row to 7B, add an exact-row documentation regression, and align
+  existing documentation tests with the newly added model row and pinned Diffusers installation.
+- **Lesson**: A versioned model name does not imply architectural continuity; verify parameter
+  counts against the official model card and lock the specific table row rather than a global count.
+- **Related Constraint**: N/A
+
 ## Cross-refs
 
 - UP: [Hard Constraints](../constraints.md), [Architecture](../architecture.md)

README.md

@@ -82,7 +82,7 @@ This experimental feature leverages `diffusers`'s `transformer.set_attention_bac
   <tr><td><a href="https://huggingface.co/ByteDance-Seed/BAGEL-7B-MoT">BAGEL-7B-MoT</a></td><td>14B</td><td>bagel</td></tr>
   <tr><td><a href="https://huggingface.co/sensenova/SenseNova-U1-8B-MoT">SenseNova-U1 1.0</a></td><td>16B</td><td>sensenova</td></tr>
   <tr><td><a href="https://huggingface.co/sensenova/SenseNova-U1.5-8B-MoT">SenseNova-U1 1.5</a></td><td>16B</td><td>sensenova</td></tr>
-  <tr><td><a href="https://huggingface.co/Qwen/Qwen-Image-2.1">Qwen-Image 2.1</a></td><td>20B</td><td>qwen-image-2.1</td></tr>
+  <tr><td><a href="https://huggingface.co/Qwen/Qwen-Image-2.1">Qwen-Image 2.1</a></td><td>7B</td><td>qwen-image-2.1</td></tr>
 
   <tr><td rowspan="4">Text-to-Video</td><td><a href="https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B-Diffusers">Wan2.1-T2V-1.3B</a></td><td>1.3B</td><td>wan2_t2v</td></tr>
   <tr><td><a href="https://huggingface.co/Wan-AI/Wan2.1-T2V-14B-Diffusers">Wan2.1-T2V-14B</a></td><td>14B</td><td>wan2_t2v</td></tr>

tests/docs/test_minimax_h3_docs.py

@@ -141,13 +141,16 @@ def test_gpu_validation_matrix_declares_scope_and_completed_result() -> None:
     assert "| Wan first/last | `Wan2.2-I2V-A14B-Diffusers` |" in text
 
 
-def test_install_docs_use_the_released_diffusers_runtime() -> None:
+def test_install_docs_distinguish_released_and_pinned_diffusers_runtimes() -> None:
     readme = _text("README.md")
     dockerfile = _text("docker/docker-cuda/Dockerfile")
     docker_readme = _text("docker/README.md")
+    qwen_news = readme[readme.index("* **[2026-09-20]**") : readme.index("* **[2026-08-21]**")]
+    h3_news = readme[readme.index("* **[2026-08-21]**") : readme.index("* **[2026-04-25]**")]
 
     assert "diffusers>=0.40.0" in readme
-    assert "pip install -e ./diffusers" not in readme
+    assert "pip install -e ./diffusers" in qwen_news
+    assert "pip install -e ./diffusers" not in h3_news
     assert "pip install -e ./diffusers" not in dockerfile
     assert "submodule (required)" not in docker_readme
 

tests/docs/test_qwen_image_21_docs.py

@@ -0,0 +1,28 @@
+# Copyright 2026 Jayce-Ping
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[2]
+
+
+def test_readme_lists_qwen_image_21_as_7b() -> None:
+    readme = (ROOT / "README.md").read_text(encoding="utf-8")
+    model_link = '<a href="https://huggingface.co/Qwen/Qwen-Image-2.1">Qwen-Image 2.1</a>'
+    row_start = readme.index(model_link)
+    model_row = readme[row_start : readme.index("</tr>", row_start)]
+
+    assert "<td>7B</td>" in model_row
+    assert "<td>20B</td>" not in model_row
+    assert "<td>qwen-image-2.1</td>" in model_row

tests/docs/test_sensenova_docs.py

@@ -24,7 +24,7 @@ def _text(path: str) -> str:
 def test_main_readme_lists_sensenova_as_t2i_and_multi_reference_i2i() -> None:
     readme = _text("README.md")
     t2i_start = readme.index('<tr><td rowspan="6">Text-to-Image</td>')
-    combined_start = readme.index('<tr><td rowspan="8">Text-to-Image & Image(s)-to-Image</td>')
+    combined_start = readme.index('<tr><td rowspan="9">Text-to-Image & Image(s)-to-Image</td>')
     video_start = readme.index('<tr><td rowspan="4">Text-to-Video</td>')
 
     t2i_only_rows = readme[t2i_start:combined_start]